What does the BLISS Science pillar measure?
—AI methods showing up in other people’s fields.
AWAITING BASELINE · 2 sources collecting, not yet scored
BLISS
What is BLISS, the SIREN upside index?
The other ending.
Every AI index measures how bad. This one measures the other direction, with the same machinery: five pillars, the same percentile normalisation against a frozen reference, the same smoothing, the same anti-flap rules. SIREN counts takeover-adjacent tempo. BLISS counts science, medicine, access, adoption and openness. Neither is a prediction.
The pairing is the instrument. Two numbers computed by identical code over different inputs can be subtracted. Two numbers produced by human or model judgement cannot. That is the whole argument for publishing both.
How do BLISS and SIREN sit on one axis?
BLISS
upside tempo
NO SCORE · awaiting a frozen baseline
SIREN
takeover-adjacent tempo
56.6
SIREN 4 · ROUTINE
The two indices cannot be compared in this build.
BLISS has no frozen reference distribution yet, so its sources are collected and published but not scored. Both scores are percentile positions against each index's own frozen reference, so the subtraction is only meaningful when both exist. There is no number here rather than a zero, because a zero is a reading somebody could have earned and this is the absence of one.
Both axes are the same 0–100 scale with the same band floors at 35, 55, 70 and 85. A score is a percentile position against that index's own frozen reference distribution — never a probability, and never a comparison between the two fields.
Why is BLISS awaiting its baseline?
10 of 10 sources are reporting and 10 of them have no frozen reference distribution to be scored against yet. A score is a percentile against a frozen record of how this source normally behaves; until that record exists there is no percentile to compute, so there is no number here.
This is the same state SIREN's markets pillar sits in, and it is deliberately not the same thing as dark. Dark means a pipe is dead. Awaiting baseline means the measurement is arriving and the ruler has not been built yet. Every value the sources returned is published below, unrounded and with its unit, so the record being accumulated is visible while it accumulates.
Nothing on this page is imputed, back-filled or estimated in the meantime. The alternative — printing a plausible number over an empty reference — is the failure this project was built in reaction to.
What are the five BLISS pillars?
What does the BLISS Medicine pillar measure?
—Trials started, clinical results indexed.
AWAITING BASELINE · 2 sources collecting, not yet scored
What does the BLISS Access pillar measure?
—What a dollar buys, and what you may keep.
AWAITING BASELINE · 2 sources collecting, not yet scored
What does the BLISS Adoption pillar measure?
—People building, not labs announcing.
AWAITING BASELINE · 2 sources collecting, not yet scored
What does the BLISS Openness pillar measure?
—How much of it anybody is allowed to hold.
AWAITING BASELINE · 2 sources collecting, not yet scored
Which BLISS sources answered this run?
10 of 10 sources answered this run — 0 scored, 10 awaiting a baseline, 0 dark. Dark sources are excluded from every number on this page, never imputed and never counted as zero. Sources awaiting a baseline are collected and published in full; they simply have no frozen history to be scored against yet.
| Source | Pillar | Value | Unit | State |
|---|---|---|---|---|
| arxiv-science-ai arXiv AI-method papers outside cs.* | science | 102 | papers/7d | awaiting baseline collected, published, not yet scored |
| openalex-ai-science OpenAlex AI-in-science articles | science | 76 | works/30d | awaiting baseline collected, published, not yet scored |
| clinicaltrials-ai ClinicalTrials.gov AI trial starts | medicine | 733 | trials/180d | awaiting baseline collected, published, not yet scored |
| pubmed-ai-clinical PubMed AI-in-clinical-practice records | medicine | 1,963 | records/30d | awaiting baseline collected, published, not yet scored |
| hf-permissive-models Open-weight models under a permissive licence | access | 1,638 | models/24h | awaiting baseline collected, published, not yet scored |
| openrouter-cost-floor Tokens per dollar at the cheap frontier | access | 8.16 | Mtok/usd | awaiting baseline collected, published, not yet scored |
| hf-model-downloads Downloads of widely deployed open models | adoption | 242,762,160 | downloads/30d | awaiting baseline collected, published, not yet scored |
| npm-ai-installs npm installs of the AI application stack | adoption | 130,861,059 | installs/7d | awaiting baseline collected, published, not yet scored |
| epoch-open-weights Open-weight share of notable models | openness | 0.3818 | share/365d | awaiting baseline collected, published, not yet scored |
| github-permissive-ai New permissively licensed ML repositories | openness | 294 | repos/30d | awaiting baseline collected, published, not yet scored |
How is BLISS computed?
- Normalise. Each source value becomes an empirical percentile against a
frozen reference distribution, then a pseudo-z through the inverse normal CDF,
then
S = 50 + 12.5·z, clamped to 0–100. - Smooth. EPA NowCast over the trailing 12 observations, weight floor 0.5.
- Pillar. The mean of its live sources' smoothed scores. A pillar with no live source is dark; a pillar whose sources have no baseline yet is awaiting one. These are different states and are never merged.
- Composite.
0.7·mean(pillars) + 0.3·max(pillars), over live pillars only. - Level. BLISS 5–1 from the composite, through the six-layer anti-flap state machine: Schmitt deadband of 3 points, dual-window dwell, a minimum interval between changes, a post-change lock, one step per change, and a 2-of-5 pillar quorum. A level change is frozen while any pillar is dark.
Every function in that list is imported from collector/engine.mjs
— the same code that computes SIREN, not a second implementation of it.
Two implementations that could disagree would destroy the only claim that makes the pairing
worth anything, which is that the two numbers are commensurable.
Engine 1.0.0 · BLISS 1.0.0 · reference not yet built · compiled .
How the SIREN AI activity index is computed · methodology AI activity index today · SIREN
What questions do people actually ask?
What is BLISS, the SIREN upside index?
Every AI index measures how bad. This one measures the other direction, with the same machinery: five pillars, the same percentile normalisation against a frozen reference, the same smoothing, the same anti-flap rules. SIREN counts takeover-adjacent tempo. BLISS counts science, medicine, access, adoption and openness. Neither is a prediction.
Why publish BLISS beside the main SIREN number?
The pairing is the instrument. Two numbers computed by identical code over different inputs can be subtracted. Two numbers produced by human or model judgement cannot. That is the whole argument for publishing both.
How do BLISS and SIREN sit on one axis?
Both axes are the same 0–100 scale with the same band floors at 35, 55, 70 and 85. A score is a percentile position against that index's own frozen reference distribution — never a probability, and never a comparison between the two fields.
Why is BLISS awaiting its baseline?
10 of 10 sources are reporting and 10 of them have no frozen reference distribution to be scored against yet. Awaiting baseline means the measurement is arriving and the ruler has not been built yet — not the same thing as dark.
What are the five BLISS pillars?
BLISS pillars are science, medicine, access, adoption and openness — the same five-pillar percentile machinery as the main index, over upside inputs.
Which BLISS sources answered this run?
10 of 10 sources answered this run — 0 scored, 10 awaiting a baseline, 0 dark. Dark sources are excluded from every number on this page, never imputed and never counted as zero.
How is BLISS computed?
Every function is imported from collector/engine.mjs — the same code that computes SIREN, not a second implementation of it. Two implementations that could disagree would destroy the only claim that makes the pairing worth anything.
Is a BLISS score a probability?
A score is a percentile position against that index's own frozen reference distribution — never a probability, and never a comparison between the two fields.
Does SIREN publish a BLISS doom score?
No. Activity tempo, not probability of harm. BLISS is an upside tempo reading beside the main index — not a probability of harm and not a forecast.
What does the BLISS Science pillar measure?
AI methods showing up in other people’s fields.
What does the BLISS Medicine pillar measure?
Trials started, clinical results indexed.
What does the BLISS Access pillar measure?
What a dollar buys, and what you may keep.
What does the BLISS Adoption pillar measure?
People building, not labs announcing.
What does the BLISS Openness pillar measure?
How much of it anybody is allowed to hold.
Search-shaped sibling: What does the SIREN infrastructure index read?.